Building Extraction from UAV Images Based on Attention Enhancement

Automatic extraction of buildings from remote sensing images using deep learning methods is crucial for urban and rural construction and management. However, the existing models are affected by the background noise and the complexity of building types during the extraction process, resulting in poor building extraction. In order to solve this problem, this paper proposes an improved model based on Mask R-CNN, which establishes a building instance extraction model for UAV imagery by adding a CBAM attention mechanism module to the backbone residual neural network ResNet. The traditional villages around Beijing are selected as the study area, and the comprehensive experimental results on the homemade building dataset show that the mAP value of the improved model is 86.4%, which is 2.1% higher than that of the original Mask R-CNN model, indicating that the improved model is more effective in building extraction.